Optimal experiment design to calibrate both a linear and non-linear viscoelastic material model for human bone tissue
I'm currently working on a study to calibrate a material model for the viscoelastic behaviour of human bone. I know that it's a strech from polymeres to human tissue, but literaturewise this field is by far the richest, when it comes to viscoelastic behaviour.
I have 13 bone samples in the form of small bone disks with different bone densities. Due to the nature of their origin, they are quiet small (r = 8mm, h=5mm), and can only be tested in compression testing.
The aim of the study is to calibrate a viscoelastic material model to be used in a bigger FEA in Abaqus. The main focus of the FEA lies on the relaxiation behavior of the bone material after a hip implantation.
For this purose both a linear and non-linear viscoelastic model model should be calibrated. The linear is more as a reference to compare the current implementation to the improved one with non-linear formulation.
My first question now is what the optimal design of eperiment would be for the szenario?
I've read all the blog posts regarding that topic and watched the corresponding videos, but I'm unsure what is applicable in my case. My idea were relaxation tests at different strain rates (and strain levels?), since the main application for the model is the relaxation response. But which strain levels and at which strain rates would be well suited for a calibration with mCalibrate? (The yield strain of bone is at about 8%)
Thank you so much for your work and making your resources freely accessable. It hepled my immensly in my understanding of viscoelastic behaviour and its calibration.
P.S. furthermore the problem is, that only about 5 of the 13 bone samples have a similar bone density. The others are scattered around them.
An idea was to calibrate the material model for those 5 with 5 different strain rate/strain level relaxation tests. Then the sensitivity of the model with respect to the bone desity could be checked by applying the calibrated model onto the other samples and correlate the resulting increase in error to the density.